feat: support DSH 0.2.2-rc.2 and continue failed generations

README.md

@@ -11,7 +11,7 @@ Connect [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) to t
 Pages become structured text with a numbered inventory of interactive elements. Tools address elements by number or selector; `browser_screenshot` captures the controlled tab's viewport after approval. The side panel also accepts PNG, JPEG, WebP, and GIF attachments when the host advertises image support.
 
 > [!IMPORTANT]
-> The workspace is pinned to DSH 0.2.1-rc.5.
+> The workspace is pinned to DSH 0.2.2-rc.2.
 
 ## Quick install
 
@@ -64,12 +64,13 @@ The paired Playwright / extension duration ratio was **1.24** (95% CI **1.16–1
 | Wait for stability | `browser_wait` | Page-load and render-settle detection |
 | Send images | `session.prompt` / `session.attachment` | Host-capability-gated image drafts, image-only prompts, and durable history previews |
 | Quote a selection | side panel composer | Text you highlight in the page appears in the composer and is sent with your next message as fenced, attributed page content |
+| Continue generating | side panel | Continue an interrupted response when the host offers it |
 
 Clicks, text entry, and scrolling run in the selected document, including cross-origin iframes. Chrome uses its debugger for native keyboard events, caret insertion, system paste, and file uploads. Screenshots require an active tab in a focused window; `browser_follow_tab { activate: true }` brings the controlled page to the front.
 
 ## Detailed installation and usage
 
-Requirements: Node.js `>=24`, Corepack/pnpm, and Chrome 125+ or Firefox 140+. Windows additionally needs Windows PowerShell 5.1, which ships with Windows, or PowerShell 7+.
+Requirements: Node.js `>=24.9.0`, Corepack/pnpm, and Chrome 125+ or Firefox 140+. Windows additionally needs Windows PowerShell 5.1, which ships with Windows, or PowerShell 7+.
 
 ### Install or update
 
@@ -125,7 +126,7 @@ cd ~/.dsh/dsh-browser && pnpm start
 From a source checkout, run `pnpm start` in the repository root. The exact supported public runtime is:
 
 ```sh
-npx @deepseek-ai/dsh@0.2.1-rc.5 web
+npx @deepseek-ai/dsh@0.2.2-rc.2 web
 ```
 
 Local Chrome use requires no configuration; Firefox requires the local bridge token described above. Open a page, click the DeepSeek whale icon, and wait for **Connected**. Existing HTTP(S) tabs are instrumented on the first action. On browser-protected pages and extension stores, the model can read tab metadata and use browser-level HTTP(S) navigation, back, forward, and reload, but it cannot inspect or operate the protected page DOM.

README.zh.md

@@ -11,7 +11,7 @@
 页面会转换为结构化文本和带编号的交互元素清单。工具通过编号或选择器定位元素；`browser_screenshot` 经批准后截取受控标签页的可视区域。宿主声明图片能力时，侧栏也可发送 PNG、JPEG、WebP 和 GIF 附件。
 
 > [!IMPORTANT]
-> 工作区固定使用 DSH 0.2.1-rc.5。
+> 工作区固定使用 DSH 0.2.2-rc.2。
 
 ## 快速安装
 
@@ -64,12 +64,13 @@ Playwright / 扩展的配对耗时比为 **1.24**（95% CI **1.16–1.34**）：
 | 等待稳定 | `browser_wait` | 页面加载与渲染稳定检测 |
 | 发送图片 | `session.prompt` / `session.attachment` | 按宿主能力启用图片草稿、纯图片消息和持久历史预览 |
 | 引用选中内容 | 侧栏输入框 | 在页面里划选的文字会出现在输入框，随下一条消息一起发送，并带上来源与不可信内容边界 |
+| 继续生成 | 侧栏 | 请求失败后，点击出现的“继续生成”按钮继续对话 |
 
 点击、文本输入和滚动操作在指定文档中执行，支持跨域 iframe。Chrome 通过调试器执行原生键盘事件、光标位置插入、系统粘贴和文件上传。截图要求受控标签页处于活动状态且窗口聚焦，可以使用 `browser_follow_tab { activate: true }` 将页面切到前台。
 
 ## 详细安装与使用
 
-前置要求：Node.js `>=24`、Corepack/pnpm，以及 Chrome 125+ 或 Firefox 140+。Windows 还需要系统自带的 Windows PowerShell 5.1，或 PowerShell 7+。
+前置要求：Node.js `>=24.9.0`、Corepack/pnpm，以及 Chrome 125+ 或 Firefox 140+。Windows 还需要系统自带的 Windows PowerShell 5.1，或 PowerShell 7+。
 
 ### 安装或更新
 
@@ -125,7 +126,7 @@ cd ~/.dsh/dsh-browser && pnpm start
 使用源码 checkout 时，请在仓库根目录运行 `pnpm start`。受支持的精确公开版本为：
 
 ```sh
-npx @deepseek-ai/dsh@0.2.1-rc.5 web
+npx @deepseek-ai/dsh@0.2.2-rc.2 web
 ```
 
 Chrome 本机使用无需配置；Firefox 需要填写上述本地桥 token。打开页面，点击 DeepSeek 鲸鱼图标，等待侧边栏显示**已连接**。已有 HTTP(S) 标签页会在第一次操作时自动加载。在浏览器受保护页面和扩展商店中，模型可以读取标签页元数据，并通过浏览器级能力导航到 HTTP(S)、后退、前进和刷新，但不能读取或操作受保护页面的 DOM。

extensions/dsh-browser/src/background/transient-events.ts

@@ -8,7 +8,7 @@ type EventFrame = Extract<ServerFrame, { t: 'event' }>
  */
 export class TransientEventCache {
   private readonly questions = new Map<string, EventFrame>()
-  private readonly running = new Map<string, EventFrame>()
+  private readonly states = new Map<string, EventFrame>()
 
   ingest(frame: ServerFrame): void {
     if (frame.t !== 'event' || !isRecord(frame.frame.payload)) return
@@ -26,21 +26,22 @@ export class TransientEventCache {
       }
       return
     }
+    if (frame.frame.method === 'session/state') {
+      this.states.set(sessionId, frame)
+      return
+    }
     if (frame.frame.method !== 'session/event' || !isRecord(payload.event)) return
-    if (payload.event.type === 'turn/start') {
-      this.running.set(sessionId, frame)
-    } else if (payload.event.type === 'turn/end') {
-      this.running.delete(sessionId)
+    if (payload.event.type === 'turn/end') {
       this.deleteSessionQuestions(sessionId)
     }
   }
 
   replay(): EventFrame[] {
-    return [...this.running.values(), ...this.questions.values()]
+    return [...this.states.values(), ...this.questions.values()]
   }
 
   clear(): void {
-    this.running.clear()
+    this.states.clear()
     this.questions.clear()
   }